Physical Layer & MAC Layer Architecture in RustV2XSimulator

This document provides a comprehensive technical overview and deep dive into the Physical (PHY) Layer and Medium Access Control (MAC) Layer implementations in rustv2xsimulator.

---

1. High-Level Architecture & Simulation Lifecycle

The simulator is built around a deterministic, discrete-event simulation (DES) engine (DiscreteEventScheduler) operating with a 100-nanosecond tick resolution (SimTime). The MAC and PHY layers work in tight coordination with the vehicular mobility engine and the ETSI C-ITS protocol stack (rustflexstack).

sequenceDiagram
    autonumber
    participant App as C-ITS Services (CAM / DENM)
    participant TxLink as Sender SimulatorLinkLayer
    participant MAC as EdcaChannelAccess
    participant Engine as SimulationEngine / Scheduler
    participant PHY as RadioMedium (Rayon)
    participant RxMAC as Receiver ChannelMonitor
    participant RxLink as Receiver SimulatorLinkLayer

    Note over App,RxLink: 1. Transmission Pipeline
    App->>TxLink: Enqueue packet (CAM / DENM)
    TxLink->>MAC: enqueue_packet(AccessCategory, TxFrame)
    MAC->>MAC: Internal contention check & Backoff calculation
    MAC->>Engine: Schedule MacTxStart (now + backoff_delay)

    Note over Engine,PHY: 2. Air Transmission & Channel Evaluation
    Engine->>MAC: Event: MacTxStart
    Engine->>Engine: Query KD-Tree (Spatial Index, radius 800m)
    Engine->>RxMAC: Record channel busy interval [now, now + duration]
    Engine->>PHY: Register AirPacket in RadioMedium
    Engine->>PHY: evaluate_broadcast(packet, sender_pos, candidates, buildings)
    PHY->>PHY: Parallel Rayon: Path Loss + Obstruction + Diffraction + Chunked SINR
    PHY-->>Engine: Vec<ReceptionResult> (is_success, rx_power, sinr)
    Engine->>Engine: Schedule PhyRxDelivery (now + duration + prop_delay)

    Note over Engine,RxLink: 3. Reception Pipeline
    Engine->>RxLink: Event: PhyRxDelivery (if is_success == true)
    RxLink->>App: Forward to GeoNetworking -> BTP -> Facility (LDM / EEBL)

---

2. MAC Layer (src/mac/)

The MAC layer is implemented in src/mac/ and provides:

1. IEEE 802.11p EDCA (Enhanced Distributed Channel Access) with 4 Access Categories (ACs).
2. Internal contention resolution between AC queues within the same node.
3. Backoff delay computation and slot-based CSMA/CA countdown.
4. Channel Busy Ratio (CBR) monitoring over sliding observation windows for Decentralized Congestion Control (DCC).
5. Unslotted ALOHA mode option for ablation studies and baseline comparisons.

2.1. Access Categories & Contention Parameters

In IEEE 802.11p (10 MHz channel bandwidth in the 5.9 GHz ITS band), medium access priority is enforced by tailoring the Arbitration Inter-Frame Space Number (\(AIFSN\)) and the Contention Window (\(CW\)) limits for each Access Category:

Access Category	Identifier	Target Traffic Type	\(AIFSN\)	\(CW_{min}\)	\(CW_{max}\)
VoiceSafety	AC_VO (0)	Emergency / High-Priority Safety (DENM, EEBL)	2	3	7
Video	AC_VI (1)	High-throughput video / perception sharing	3	7	15
BestEffort	AC_BE (2)	Standard periodic Cooperative Awareness (CAM)	6	15	1023
Background	AC_BK (3)	Non-safety background telemetry	9	15	1023

Source Reference: EdcaQueueConfig::for_ac.

2.2. Timing Parameters & Frame Transmission Duration

• Slot Time (\(\sigma\)): \(13\,\mu\text{s}\) (slot_time)
• SIFS (Short Interframe Space): \(32\,\mu\text{s}\) (sifs)
• CCA Threshold: \(-85.0\,\text{dBm}\)

Arbitration Inter-Frame Space (\(AIFS\))

\[AIFS[AC] = SIFS + AIFSN[AC] \times \sigma\]

Backoff Contention Delay

When a frame is queued for transmission, a uniform random backoff is selected:

\[\text{backoff\_slots} \in [0, CW_{min}]\]

\[\text{Delay}_{contention} = AIFS[AC] + \text{backoff\_slots} \times \sigma\]

802.11p Frame Duration on Air

For an OFDM transmission at 6 Mbps with QPSK rate-1/2 over a 10 MHz channel (\(N_{DBPS} = 24\) data bits per OFDM symbol of \(8\,\mu\text{s}\) duration), the packet duration is calculated in compute_duration_10mhz:

\[\text{duration} = 40\,\mu\text{s} + \left\lceil \frac{16 + 8 \times L_{bytes} + 6}{24} \right\rceil \times 8\,\mu\text{s}\]

(where \(40\,\mu\text{s}\) accounts for the physical preamble and PLCP header, 16 bits for SERVICE field, and 6 tail bits).

2.3. Internal Contention Resolution

Each vehicle node owns an EdcaChannelAccess with four queues (EdcaQueue). When multiple queues have pending frames:

1. poll_ready_transmission evaluates queues in descending order of priority:

$\(\text{VoiceSafety} > \text{Video} > \text{BestEffort} > \text{Background}\)$

1. The winning queue dequeues its frame, resets its contention window \(CW \leftarrow CW_{min}\), and transitions to MacState::Idle.
2. The remaining subordinate queues with pending traffic encounter an internal collision:

$\(CW \leftarrow \min(2 \times CW + 1, CW_{max})\)$

and draw a new random backoff counter (on_internal_collision).

2.4. Channel Monitor & Channel Busy Ratio (CBR)

The ChannelMonitor tracks local channel utilization over a sliding window \(T_{CBR} = 100\,\text{ms}\):

1. Interval Tracking: When any transmission starts in transmission range, record_busy_interval(start, end) logs the active period.
2. Window Trimming: In evaluate_cbr(now), all historical intervals ending prior to \(\text{now} - T_{CBR}\) are pruned.
3. Overlapping Interval Union: Because multiple transmissions may overlap in time, intervals are sorted and merged into disjoint busy segments:

$\(T_{busy} = \bigcup_{k} [start_k, end_k] \cap [now - T_{CBR}, now]\)$

1. CBR Ratio:

$\(\text{CBR} = \frac{T_{busy}}{T_{CBR}} \in [0.0, 1.0]\)$

1. DCC Interaction: Every 100 ms, the simulation engine triggers EventPayload::CbrUpdate, which queries evaluate_cbr() and feeds it into the vehicle's Decentralized Congestion Control (either ReactiveDccFsm or LimericDcc) to throttle CAM transmission intervals (\(T_{off}\)).

---

3. Physical (PHY) Layer (src/phy/)

The physical layer resides in src/phy/. It manages radio wave propagation, obstacle attenuation, corner diffraction, cumulative interference, capture effects, and frame error decoding.

flowchart TD
    AirPacket["AirPacket (Tx Power, Duration, Start Time)"]
    KDTree["KD-Tree Query: Candidates within 800m"]
    RayonPar["Rayon Parallel Loop (Per Candidate Receiver)"]

    BuildingObs["BuildingMap Raycasting<br/>- Wall Crossings<br/>- Penetration Distance (0.85 dB/m)"]
    PathLoss["Dual-Slope Log-Distance<br/>- Breakpoint: 120m<br/>- gamma1 = 2.05, gamma2 = 3.90<br/>- Shadow Fading: sigma_LoS=3.1dB, sigma_NLoS=8.3dB"]
    KnifeEdge["ITU-R P.526 Knife-Edge Corner Diffraction<br/>- Closest occluding corner<br/>- Fresnel parameter nu<br/>- J(nu) attenuation (0 to 45 dB)"]

    RxPower["Received Signal Power: P_rx = P_tx - PL_total"]

    InterfTracker["CumulativeInterferenceTracker (ns-3 Parity)<br/>- Piecewise Constant Temporal Chunks<br/>- Cumulative Interference Sum I(t)<br/>- Noise Floor: -99 dBm"]
    Capture["Capture Effect Check<br/>- Margin: gamma_cap = 4.0 dB<br/>- SIR < 4.0 dB & SINR < 6.0 dB -> Collision"]
    PER["Analytical PER Curve (802.11p QPSK 1/2 6 Mbps)<br/>- NistErrorRateModel<br/>- Multi-chunk PSR Product: PSR = prod(1 - PER_i)"]
    MonteCarlo["Deterministic Monte Carlo Decision<br/>rand < PSR -> is_success"]

    AirPacket --> KDTree --> RayonPar
    RayonPar --> BuildingObs
    BuildingObs --> PathLoss
    BuildingObs --> KnifeEdge
    PathLoss & BuildingObs & KnifeEdge --> RxPower
    RxPower --> InterfTracker
    InterfTracker --> Capture
    Capture --> PER --> MonteCarlo

3.1. Parallel Evaluation & Spatial Filtering

When handle_mac_tx fires:

1. The engine constructs an AirPacket (\(P_{tx} = 23\,\text{dBm}\), standard duration \(\sim 300\,\mu\text{s}\)).
2. The 2D spatial index (SpatialIndex) finds all candidate receivers within a cutoff radius of \(800\,\text{m}\).
3. All candidates are evaluated concurrently using rayon::prelude::*.
4. Each receiver utilizes a deterministic pseudo-random generator keyed to the simulation seed, packet ID, receiver ID, and iteration index:

   let mut rng = StdRng::seed_from_u64(
       sim_seed ^ ((packet.packet_id << 32) | (rx_id as u64)) ^ (idx as u64)
   );
   

---

3.2. Propagation and Path Loss Models

The simulator offers two selectable PHY models (PhyModelType):

1. Dual-Slope Log-Distance Model (PhyModelType::DualSlope, Default)

Implemented in DualSlopePathLoss:

• Carrier Frequency: \(f_c = 5.9\,\text{GHz}\).
• Reference Path Loss at \(d_0 = 1.0\,\text{m}\):

$\(PL_0 = 20 \log_{10}\left( \frac{4 \pi \cdot 1.0 \cdot f_c}{c} \right) \approx 47.85\,\text{dB}\)$

• Breakpoint Distance (\(d_{break}\)): \(120.0\,\text{m}\).
• Path Loss Exponents: \(\gamma_1 = 2.05\) (before breakpoint), \(\gamma_2 = 3.90\) (after breakpoint).

Deterministic Path Loss:

\[PL_{det}(d) = \begin{cases} PL_0 + 10 \gamma_1 \log_{10}(d), & d \le d_{break} \\ PL_0 + 10 \gamma_1 \log_{10}(d_{break}) + 10 \gamma_2 \log_{10}\left(\frac{d}{d_{break}}\right), & d > d_{break} \end{cases}\]

Log-Normal Shadow Fading (\(X_\sigma\)): A zero-mean Gaussian random variable sampled according to Line-of-Sight conditions:

\[X_\sigma \sim \mathcal{N}(0, \sigma^2)\]

• Line-of-Sight (\(\text{LoS}\)): \(\sigma_{LoS} = 3.1\,\text{dB}\)
• Non-Line-of-Sight (\(\text{NLoS}\)): \(\sigma_{NLoS} = 8.3\,\text{dB}\)

---

2. Building Obstruction & Wall Penetration Attenuation

Implemented in BuildingMap::evaluate_obstruction:

• Intersects the transmitter-to-receiver 2D ray against all building polygons parsed from SUMO/OSM (osm.poly.xml).
• Calculates cumulative internal penetration distance (\(d_{pen}\)).
• Computes concrete attenuation:

$\(PL_{penetration} = \min(d_{pen} \times 0.85\,\text{dB/m},\, 45.0\,\text{dB})\)$

• If any building is intersected, \(\text{LoS}\) is set to false.

---

3. ITU-R P.526 Knife-Edge Corner Diffraction

Implemented in FresnelKnifeEdge: When a building obstructs the direct line of sight, the simulator extracts the closest corner vertex of the building to the ray and calculates continuous corner diffraction:

1. Fresnel-Kirchhoff dimensionless parameter \(\nu\):

$\(\nu = h \sqrt{\frac{2(d_1 + d_2)}{\lambda d_1 d_2}}\)$

(where \(h\) is the clearance distance, \(\lambda = c / f_c \approx 0.0508\,\text{m}\), and \(d_1, d_2\) are distances from TX and RX to the corner vertex). 2. Diffraction Loss \(J(\nu)\):

$\(J(\nu) = \begin{cases} 0\,\text{dB}, & \nu \le -0.78 \\ \left[ 6.9 + 20 \log_{10}\left( \sqrt{(\nu - 0.1)^2 + 1} + \nu - 0.1 \right) \right]_{0}^{45\,\text{dB}}, & \nu > -0.78 \end{cases}\)$

Total Received Power:

\[P_{rx} = P_{tx} - \Big( PL_{det}(d) + X_\sigma + PL_{penetration} + J(\nu) \Big)\]

---

4. Legacy Polynomial Model (PhyModelType::LegacyPoly, Baseline)

Implemented in LegacyPolynomialPathLoss for ablation study \(A4\):

• Uses an empirical 10th-degree polynomial to evaluate distance loss:

$\(P_{loss, dist}(d) = \sum_{k=0}^{10} C_k d^k \quad (\text{clamped to } [0, 1])\)$

• Combines distance loss with a logistic function modeling channel congestion based on local bytes on air:

$\(P_{loss, cong}(B) = \frac{1}{1 + e^{-k (B - b_0)}}\)$

• Delivery probability: \(P_{deliver} = (1 - P_{loss, dist}) \times (1 - P_{loss, cong})\) if LoS, else 0.

---

3.3. Chunked Cumulative SINR & Receiver Capture Effect

Implemented in CumulativeInterferenceTracker, closely mirroring the ns-3 v3.42 InterferenceHelper architecture.

gantt
    title Piecewise-Constant Interference Chunking
    dateFormat X
    axisFormat %s ms
    section Desired Signal
    Desired Packet (P_desired) :active, d1, 0, 10
    section Interferer A
    Interferer A (P_intf_A)    :crit, a1, 0, 4
    section Interferer B
    Interferer B (P_intf_B)    :crit, b1, 3, 7
    section Resulting Chunks
    Chunk 1 (0-3 ms, I = P_A)            :done, c1, 0, 3
    Chunk 2 (3-4 ms, I = P_A + P_B)      :crit, c2, 3, 4
    Chunk 3 (4-7 ms, I = P_B)            :done, c3, 4, 7
    Chunk 4 (7-10 ms, Noise only)        :active, c4, 7, 10

1. Piecewise Constant Slicing

When a packet reception occurs:

1. The tracker collects all boundary timestamps (start and end times) of concurrent transmissions active during \([t_{start}, t_{end}]\).
2. It divides the packet reception interval into contiguous time slices \([t_i, t_{i+1}]\) of duration \(\Delta t_i\).
3. For each slice \(i\), the total interfering power is summed in linear Watts:

$\(I_i = \sum_{j \in \text{interferers}} P_{rx, j}(t_{mid})\)$

$\(\text{SINR}_i = \frac{P_{rx, desired}}{I_i + N_0}\)$

(where thermal noise floor \(N_0 = -99.0\,\text{dBm} \approx 1.259 \times 10^{-13}\,\text{W}\)).

2. Receiver Capture Effect (\(\gamma_{cap} = 4.0\,\text{dB}\))

If interfering signals overlap with the desired packet:

• The Signal-to-Interference Ratio is computed:

$\(\text{SIR}_i = \frac{P_{rx, desired}}{I_i}\)$

• Capture condition: If \(\text{SIR}_i < 4.0\,\text{dB}\) and \(\text{SINR}_i < 6.0\,\text{dB}\), destructive collision occurs and reception fails immediately (\(\text{PSR} = 0\)).
• If \(\text{SIR}_i \ge 4.0\,\text{dB}\), the receiver successfully captures the stronger desired packet over the interferer.

3. Analytical Packet Error Rate (PER) Curve

For each chunk, the Packet Error Rate for IEEE 802.11p OFDM 6 Mbps (QPSK 1/2 with convolutional coding) is computed using the ns-3 NistErrorRateModel logistic curve in compute_80211p_coded_per:

\[\text{PER}_i(\text{SINR}_i) = \begin{cases} 1.0, & \text{SINR}_{i,\text{dB}} < 2.0\,\text{dB} \\ 0.0, & \text{SINR}_{i,\text{dB}} > 12.0\,\text{dB} \\ \min\left(1.0, \, \frac{1}{1 + \exp\left(1.6 (\text{SINR}_{i,\text{dB}} - 5.5)\right)} \times \text{clamp}\left(\frac{L_{bytes}}{300}, 0.5, 3.0\right)\right), & 2.0 \le \text{SINR}_{i,\text{dB}} \le 12.0 \end{cases}\]

4. Total Packet Success Rate (PSR)

The probability of receiving the entire packet successfully is the product of success probabilities across all time chunks:

\[\text{PSR} = \prod_{i} (1 - \text{PER}_i)\]

Finally, a random float \(u \sim \mathcal{U}(0, 1)\) determines whether the reception succeeds (\(u < \text{PSR}\)).

---

4. Verification & ns-3 Parity

The PHY layer has been validated against an ns-3 v3.42 gold standard in tests/ns3_phy_parity_test.rs:

• Packet Delivery Ratio (PDR) is evaluated across \(10\,\text{m}\) distance bins up to \(800\,\text{m}\).
• Mean Absolute Error (MAE) against ns-3 is asserted to satisfy:

$\(\text{MAE}_{\text{PDR}} \le 0.04\)$

• The integration test verifies that the dual-slope breakpoint, log-normal shadowing, and analytical error models match standard ns-3 link behavior with high fidelity.

---

5. Summary Table: MAC vs. PHY Layer Responsibilities

Dimension	MAC Layer (src/mac/)	PHY Layer (src/phy/)
Primary Module	EdcaChannelAccess & ChannelMonitor	RadioMedium & CumulativeInterferenceTracker
Standards	IEEE 802.11p EDCA, ETSI TS 102 687 (DCC)	IEEE 802.11p PHY, ITU-R P.526, ns-3 v3.42
QoS / Prioritization	4 Access Categories (AC_VO, AC_VI, AC_BE, AC_BK)	Equal treatment at RF; prioritized by TX duration & backoff
Collision Resolution	Internal contention (CW doubling), Backoff delay	Receiver capture effect (\(\gamma_{cap} = 4\,\text{dB}\)), SINR degradation
Channel Evaluation	CBR over sliding \(100\,\text{ms}\) window	Path loss, shadow fading, building penetration, diffraction
Concurrency	Per-agent state machines, single-threaded DES dispatch	Parallel multi-receiver broadcast evaluation using Rayon
Event Triggers	Schedules MacTxStart when backoff completes	Schedules PhyRxDelivery at duration + propagation delay